A large-scale pragmatic automated audit of the appropriateness of antibiotics initiated for presumed community-acquired pneumonia using a novel algorithmic approach

Audit is a fundamental part of antimicrobial stewardship, but this has traditionally been labour-intensive. The advent of fully electronic records (EPIC) at Cambridge University Hospitals (CUH) presents novel opportunities for large-scale automated data analyses and feedback. We developed and validated an algorithm to audit the appropriateness of prescriptions initiated for presumed community-acquired pneumonia (CAP).

Methods

We developed an algorithm that extracts prescription and clinical data from EPIC, calculates CURB-65 scores, and assesses the appropriateness of antibiotics with an indication of CAP against Trust guidelines based on predefined rules. Clinical data included age, gender, blood results, vital signs, NEWS-2 score, MRSA status, penicillin allergy and pregnancy status. Prescriptions were limited to 48 hours from admission. The accuracy of the algorithm was validated in a representative sample of 30 patients.

We present data on all prescriptions initiated for CAP admitted to CUH between September 2018 and June 2019.

On validation, the algorithm calculated the CURB-65 score with an accuracy of 97% and correctly categorised antibiotic appropriateness in 98.5% of cases. Only 15% of patients had a CURB-65 score documented in the notes.

The algorithm evaluated 4,307 prescriptions in 2,198 patients. Appropriateness was significantly better in CURB-65 scores of 2-5 (83.7%) versus 0-1 (33.5%) largely due to over-prescription of co-amoxiclav in the latter.

This algorithm enables large-scale analysis of prescriptions initiated for CAP with high accuracy automating the audit cycle. An automatically calculated CURB-65 score has the potential to reduce over-prescribing of co-amoxiclav and should be evaluated in the future.